报告题目：3D Visual Localization in Urban Environments

报告时间：2024年5月27日  周一 15：00

报告地点：信息楼133会议室

主持人：殷赵霞 教授

报告人：夏彦  Research Fellow 德国慕尼黑工业大学

个人简介：

Dr. Yan Xia is now a Postdoctoral Researcher in the Chair of Computer Vision and Artificial Intelligence at Technical University of Munich (TUM), working with Prof. Daniel Cremers (Gottfried Wilhelm Leibniz-Prize Winner) on advancing 3D scene understanding. He is also an associate member of Munich Center for Machine Learning. He has served as secretary of Working Group I/8 International Society of Photogrammetry and Remote Sensing (ISPRS). He obtained his PhD degree from TUM in 2023 and was a visiting scholar in Visual Geometry Group (VGG) at University of Oxford. His research interests include 3D vision, robotics, and autonomous driving. He has published 20 academic papers at top-tier conferences and journals such as CVPR, ICCV and ISPRS Journal. He received the Best Paper Award at the ISPRS Geospatial Week 2023.

报告摘要：

Localization is a critical capability for self-driving vehicles, allowing them to pinpoint their location on a map. The well-known technique for localization is to use global navigation satellite system (GNSS). However, the signals of satellites are bound to fail among tall buildings and vegetation. As a result, sensor-based solutions can be decisive in improving the localization. In my research, I develop a series of 3D visual localization methods, aiming to enhance localization accuracy and generalizability in GNSS-denied environments. In this talk, I will specifically highlight several of my recent papers. These include studies on point cloud based place recognition ^[1][2] and cross-model localization methods ^[3].

[1] Xia et al.,SOE-Net: A self-attention and orientation encoding network for point cloud based place recognition. CVPR 2021 (Oral)

[2]Xia et al.,Casspr: Cross attention single scan place recognition. ICCV 2023

[3]Xia et al., Text2loc: 3d point cloud localization from natural language. CVPR 2024.